Rank-at-most-two spectral slice of a symmetric nonnegative matrix #
A symmetric entrywise-nonnegative matrix B has a rank-at-most-two PSD slice
X = a uuᵀ + b vvᵀ built from a nonnegative Perron vector for λ_max and,
when λ₂ > 0, a unit eigenvector for λ₂ orthogonal to it. This slice
satisfies ⟨B, X⟩ = ⟨X, X⟩ = F B and is a Gram matrix of planar vectors in
the closed right half-plane. If B is zero-diagonal and supported on E(G),
the Motzkin–Straus bound on M X yields F B ≤ turanFactor G * ⟨B, B⟩.
Index of λ_max among eigenvalues.
Equations
Instances For
Index of λ₂ among eigenvalues.
Equations
Instances For
A unit eigenvector for λ₂ orthogonal to a unit λ_max-eigenvector, when λ₂ > 0.
Data of the rank-at-most-two spectral slice of a symmetric nonnegative matrix.
- a : ℝ
The largest eigenvalue
λ₁(B). - b : ℝ
The positive part
max (λ₂(B)) 0of the second eigenvalue. - u : n → ℝ
A nonnegative unit eigenvector for
a. - v : n → ℝ
Instances For
SP04. Existence of the spectral slice.
Equations
- One or more equations did not get rendered due to their size.
Instances For
SP04. The spectral slice is positive semidefinite.
SP04. The spectral slice has rank at most two.
SP05. ⟨B, X⟩ = ⟨X, X⟩ = F B.
SP05. If B ≠ 0 then the common Frobenius mass is positive.
Planar Gram embedding of the spectral slice.
Instances For
SP06. The first coordinate is nonnegative.
SP06. X is the Gram matrix of the planar embedding.
SP07. M of the spectral slice is completely positive.
SP08. Completely positive Motzkin–Straus on M s.X.
SP09–SP10 — support bound and Frobenius Cauchy–Schwarz #
SP09. If B is supported on the edges of G and entrywise nonnegative,
then pairing against X is dominated by pairing against A_G ⊙ X₊.
SP10. Cauchy–Schwarz for the Frobenius pairing against A_G ⊙ X₊.
SP11–SP12 — algebra of the weighted inequality #
SP11. Combining the support bound, Cauchy–Schwarz, and SP08.
SP12. Theorem thm:weighted.